Re: [PATCH] t/lib-git.sh: fix ACL-related permissions failure

From: Jeff King <peff@peff.net>
Date: Nov 5, 2021, 23:53 UTC
Message-ID: <YYXD3NESdiDI4B6G@coredump.intra.peff.net>
In-Reply-To: <CA+kUOa=vqFNXe2QKc8K31OLL0zkEsK7wAk6hPMxjQJNVM7PsGQ@mail.gmail.com>

On Fri, Nov 05, 2021 at 06:49:14PM +0000, Adam Dinwoodie wrote:

Show 11 quoted lines

> This is probably a much broader conversation. I remember when I first
> started packaging Git for Cygwin, I produced a release that didn't
> have support for HTTPS URLs due to a missing dependency in my build
> environment. The build and test suite all passed -- it assumed I just
> wanted to build a release that didn't have HTTPS support -- so some
> relatively critical function was silently skipped. I don't know how to
> avoid that sort of issue other than relying on (a) user bug (or at
> least missing function) reports and (b) folk building Git for
> themselves/others periodically going through the output of the
> configure scripts and the skipped subtests to make sure only expected
> things get missed; neither of those options seem great to me.

The HTTP tests in particular have a knob for this, as I was worried about this kind of situation when we introduced auto-enabling of network tests back in 83d842dc8c (tests: turn on network daemon tests by default, 2014-02-10). The solution there was to make the knob a tri-state: the default is "auto", which will try to probe whether we have a working apache setup, but setting it to "true" will complain if that setup fails.

Now that's not a perfect solution:

  - you have to know to flip the switch to "true". For an old switch
    like HTTP, that's easy. But somebody packaging Git might not even
    realize GPGSSH was a new thing.

  - The "true" knob only covers probing of the environment. If you
    accidentally build with NO_CURL, we'd still quietly skip the tests.
    It might be reasonable to change this.

  - In your particular case, it probably would not have helped anyway
    because we don't have any specific HTTPS tests (there is an option
    to set up the default server with SSL, but I didn't even realize
    that until just now; I wonder if it actually works).

So I dunno. I guess because of point 1, having an allow-known-skips list would be more helpful. That gives you the opportunity to examine new prereqs and decide if they ought to be skipped or not in your setup.

-Peff